rohit/probo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add tracker description enrichment worker

Browse Source 
Tracker descriptions were only filled on the agent-identification path,
so patterns resolved by domain, sibling, or fallback stayed without one,
and empty mapping upserts could clobber a researched description on the
shared catalog row.

Move description ownership to a dedicated, global common-pattern
enrichment worker. New catalog rows are queued on insert; the worker
researches a compliance-grade description with web search, records it on
the common pattern, and fans it out to every linked tracker pattern. The
mapping worker no longer generates descriptions and only propagates an
already-enriched one at link time.

Rename TrackerMappingConfig to TrackerAgentsConfig since the mapping and
enrichment agents now share it.

Signed-off-by: Émile Ré <emile@probo.com>
This commit is contained in:
Émile Ré
2026-05-29 11:48:10 +02:00
parent 29791ae775
commit 24bece6f86
12 changed files with 665 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
pkg/probod/probod.go
View File

@@ -313,7 +313,7 @@ func (impl *Implm) Run(
return err
}
trackerMappingCfg, thirdPartyDisambiguationCfg, err := impl.buildTrackerMappingConfig(l, tp, r)
trackerAgentsCfg, thirdPartyDisambiguationCfg, err := impl.buildTrackerAgentsConfig(l, tp, r)
if err != nil {
return err
}
@@ -725,7 +725,7 @@ func (impl *Implm) Run(
},
)
trackerMappingWorker := cookiebanner.NewTrackerMappingWorker(pgClient, l, trackerMappingCfg, thirdPartyDisambiguationCfg)
trackerMappingWorker := cookiebanner.NewTrackerMappingWorker(pgClient, l, trackerAgentsCfg, thirdPartyDisambiguationCfg)
trackerMappingWorkerCtx, stopTrackerMappingWorker := context.WithCancel(context.Background())
wg.Go(
@@ -736,6 +736,26 @@ func (impl *Implm) Run(
},
)
// The common-pattern enrichment worker needs an LLM client (it
// researches descriptions via the agent), so it is only started when
// the tracker agents are configured.
stopCommonPatternEnrichmentWorker := func() {}
if trackerAgentsCfg.LLMClient != nil {
commonPatternEnrichmentWorker := cookiebanner.NewCommonPatternEnrichmentWorker(pgClient, l, trackerAgentsCfg)
var commonPatternEnrichmentWorkerCtx context.Context
commonPatternEnrichmentWorkerCtx, stopCommonPatternEnrichmentWorker = context.WithCancel(context.Background())
wg.Go(
func() {
if err := commonPatternEnrichmentWorker.Run(commonPatternEnrichmentWorkerCtx); err != nil {
cancel(fmt.Errorf("common pattern enrichment worker crashed: %w", err))
}
},
)
}
mailingListWorker := mailman.NewMailingListWorker(mailmanService, pgClient, l.Named("mailing-list-worker"))
mailingListWorkerCtx, stopMailingListWorker := context.WithCancel(context.Background())
@@ -805,6 +825,7 @@ func (impl *Implm) Run(
stopESignService()
stopTrackerPatternAnalysisWorker()
stopTrackerMappingWorker()
stopCommonPatternEnrichmentWorker()
stopMailingListWorker()
stopEvidenceDescriptionWorker()
stopDocumentPDFWorker()

27
pkg/probod/tracker_mapping.go → pkg/probod/tracker_agents.go
View File

@@ -24,25 +24,26 @@ import (
"go.probo.inc/probo/pkg/thirdparty"
)
// buildTrackerMappingConfig wires the tracker-mapping agent (catalog
// identification) and the third-party disambiguation agent that the
// tracker-mapping worker uses to promote patterns to org ThirdParties.
// Both are opt-in: deployments that do not set
// `llm.tracker-mapping.provider` get zero configs (nil LLM client) so
// the worker runs without agent fallback.
// buildTrackerAgentsConfig wires the tracker agents that share one LLM
// client and model: the tracker-mapping agent (catalog identification),
// the common-pattern enrichment agent (description research), and the
// third-party disambiguation agent that the tracker-mapping worker uses
// to promote patterns to org ThirdParties. All are opt-in: deployments
// that do not set `llm.tracker-mapping.provider` get zero configs (nil
// LLM client) so the workers run without agent fallback.
//
// Both agents are sourced from the same `tracker-mapping` config slot
// The agents are sourced from the same `tracker-mapping` config slot
// because they share the LLM client, model, and lifecycle. The
// disambiguation agent has no Firecrawl/DB tools, so its config
// surface is narrower and it lives in the cross-domain pkg/thirdparty
// package.
func (impl *Implm) buildTrackerMappingConfig(
func (impl *Implm) buildTrackerAgentsConfig(
l *log.Logger,
tp trace.TracerProvider,
r prometheus.Registerer,
) (cookiebanner.TrackerMappingConfig, thirdparty.DisambiguationConfig, error) {
) (cookiebanner.TrackerAgentsConfig, thirdparty.DisambiguationConfig, error) {
if impl.cfg.Agents.TrackerMapping.Provider == "" {
return cookiebanner.TrackerMappingConfig{}, thirdparty.DisambiguationConfig{}, nil
return cookiebanner.TrackerAgentsConfig{}, thirdparty.DisambiguationConfig{}, nil
}
agentCfg, llmClient, err := impl.resolveAgentClient(
@@ -53,10 +54,10 @@ func (impl *Implm) buildTrackerMappingConfig(
r,
)
if err != nil {
return cookiebanner.TrackerMappingConfig{}, thirdparty.DisambiguationConfig{}, fmt.Errorf("cannot resolve tracker mapping agent client: %w", err)
return cookiebanner.TrackerAgentsConfig{}, thirdparty.DisambiguationConfig{}, fmt.Errorf("cannot resolve tracker mapping agent client: %w", err)
}
mappingCfg := cookiebanner.TrackerMappingConfig{
trackerAgentsCfg := cookiebanner.TrackerAgentsConfig{
LLMClient: llmClient,
Model: agentCfg.ModelName,
FirecrawlAPIKey: impl.cfg.Agents.Tools.FirecrawlAPIKey,
@@ -67,5 +68,5 @@ func (impl *Implm) buildTrackerMappingConfig(
Model: agentCfg.ModelName,
}
return mappingCfg, disambiguationCfg, nil
return trackerAgentsCfg, disambiguationCfg, nil
}